-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathcosin_hist_-1-1_gate.py
117 lines (108 loc) · 7.95 KB
/
cosin_hist_-1-1_gate.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
dataset1 = pd.read_csv('analysis results/circular_square_cosin(-1,1)_circular_8k.csv')
dataset2 = pd.read_csv('analysis results/circular_square_cosin(-1,1)_rectangle.csv')
dataset3 = pd.read_csv('analysis results/z_circular_square_cosin(-1,1)_8k.csv')
dataset4 = pd.read_csv('analysis results/z_circular_square_cosin(-1,1)_rectangle.csv')
# 计算MMD
def gaussian_kernel(x, y, sigma=1.0):
gamma = 1.0 / (2 * sigma ** 2)
pairwise_dists = np.sum(np.square(x), axis=1).reshape(-1, 1) + np.sum(np.square(y), axis=1) - 2 * np.dot(x, y.T)
return np.exp(-gamma * pairwise_dists)
def mmd(hist_1, hist_2, sigma=1.0):
K11 = gaussian_kernel(hist_1.reshape((-1, 1)), hist_1.reshape((-1, 1)), sigma)
K22 = gaussian_kernel(hist_2.reshape((-1, 1)), hist_2.reshape((-1, 1)), sigma)
K12 = gaussian_kernel(hist_1.reshape((-1, 1)), hist_2.reshape((-1, 1)), sigma)
MMD = np.sqrt(K11.mean() + K22.mean() - 2 * K12.mean())
return MMD
# 计算KL散度
def kl_divergence(hist_1, hist_3):
# 将hist1和hist3中的0值替换为1e-10，避免出现log(0)的情况
hist_1[hist_1 == 0] = 1e-10
hist_3[hist_3 == 0] = 1e-10
# 计算KL散度
kl_div = np.sum(hist_1 * np.log(hist_1 / hist_3))
return kl_div
fig = plt.figure(figsize=(20, 20), dpi=100)
for i in range(1, 5):
fig.add_subplot(2, 2, i)
plt.subplot(221)
values1, bins1, bars1 = plt.hist(dataset1['circular_square'], edgecolor='white', bins=100, color='orange', range=(-1, 1))
hist1, _ = np.histogram(dataset1['circular_square'], bins=50, density=True)
plt.ylim(0, 450)
plt.xlabel("cosin simi(-1,1)")
plt.ylabel("Frequency")
plt.title('circular&square cosin(-1,1)_10k')
count11 = np.count_nonzero((dataset1['circular_square'] >= 0.75) & (dataset1['circular_square'] <= 1))
count12 = np.count_nonzero((dataset1['circular_square'] >= 0.5) & (dataset1['circular_square'] <= 0.75))
count13 = np.count_nonzero((dataset1['circular_square'] >= 0.25) & (dataset1['circular_square'] <= 0.5))
count14 = np.count_nonzero((dataset1['circular_square'] >= 0) & (dataset1['circular_square'] <= 0.25))
count15 = np.count_nonzero((dataset1['circular_square'] >= -0.25) & (dataset1['circular_square'] <= 0))
count16 = np.count_nonzero((dataset1['circular_square'] >= -0.5) & (dataset1['circular_square'] <= -0.25))
count17 = np.count_nonzero((dataset1['circular_square'] >= -0.75) & (dataset1['circular_square'] <= -0.5))
count18 = np.count_nonzero((dataset1['circular_square'] >= -1) & (dataset1['circular_square'] <= -0.75))
print('图1' + str((count11, count12, count13, count14, count15, count16, count17, count18)))
plt.subplot(222)
values2, bins2, bars2 = plt.hist(dataset2['circular_square'], edgecolor='white', bins=100, color='skyblue', range=(-1, 1))
hist2, _ = np.histogram(dataset2['circular_square'], bins=50, density=True)
plt.ylim(0, 350)
plt.xlabel("cosin simi(-1,1)")
plt.ylabel("Frequency")
plt.title('circular&square cosin(-1,1)_1k')
count21 = np.count_nonzero((dataset2['circular_square'] >= 0.75) & (dataset2['circular_square'] <= 1))
count22 = np.count_nonzero((dataset2['circular_square'] >= 0.5) & (dataset2['circular_square'] <= 0.75))
count23 = np.count_nonzero((dataset2['circular_square'] >= 0.25) & (dataset2['circular_square'] <= 0.5))
count24 = np.count_nonzero((dataset2['circular_square'] >= 0) & (dataset2['circular_square'] <= 0.25))
count25 = np.count_nonzero((dataset2['circular_square'] >= -0.25) & (dataset2['circular_square'] <= 0))
count26 = np.count_nonzero((dataset2['circular_square'] >= -0.5) & (dataset2['circular_square'] <= -0.25))
count27 = np.count_nonzero((dataset2['circular_square'] >= -0.75) & (dataset2['circular_square'] <= -0.5))
count28 = np.count_nonzero((dataset2['circular_square'] >= -1) & (dataset2['circular_square'] <= -0.75))
print('图2' + str((count21, count22, count23, count24, count25, count26, count27, count28)))
plt.subplot(223)
values3, bins3, bars3 = plt.hist(dataset3['z_circular_square_8k'], edgecolor='white', bins=100, color='orange', range=(-1, 1))
hist3, _ = np.histogram(dataset3['z_circular_square_8k'], bins=50, density=True)
plt.ylim(0, 450)
plt.xlabel("cosin simi(-1,1)")
plt.ylabel("Frequency")
plt.title('z square&circular cosin(-1,1)_10k')
mmd1_3 = mmd(hist1, hist3, sigma=1.0) # 计算MMD
kl1_3 = kl_divergence(hist1 / 8000, hist3 / 8000)
kl3_1 = kl_divergence(hist3 / 8000, hist1 / 8000)
count31 = np.count_nonzero((dataset3['z_circular_square_8k'] >= 0.75) & (dataset3['z_circular_square_8k'] <= 1))
count32 = np.count_nonzero((dataset3['z_circular_square_8k'] >= 0.5) & (dataset3['z_circular_square_8k'] <= 0.75))
count33 = np.count_nonzero((dataset3['z_circular_square_8k'] >= 0.25) & (dataset3['z_circular_square_8k'] <= 0.5))
count34 = np.count_nonzero((dataset3['z_circular_square_8k'] >= 0) & (dataset3['z_circular_square_8k'] <= 0.25))
count35 = np.count_nonzero((dataset3['z_circular_square_8k'] >= -0.25) & (dataset3['z_circular_square_8k'] <= 0))
count36 = np.count_nonzero((dataset3['z_circular_square_8k'] >= -0.5) & (dataset3['z_circular_square_8k'] <= -0.25))
count37 = np.count_nonzero((dataset3['z_circular_square_8k'] >= -0.75) & (dataset3['z_circular_square_8k'] <= -0.5))
count38 = np.count_nonzero((dataset3['z_circular_square_8k'] >= -1) & (dataset3['z_circular_square_8k'] <= -0.75))
print((mmd1_3, kl1_3, kl3_1))
print('图3' + str((count31, count32, count33, count34, count35, count36, count37, count38)))
plt.subplot(224)
values4, bins4, bars4 = plt.hist(dataset4['z_circular_square'], edgecolor='white', bins=100, color='skyblue', range=(-1, 1))
hist4, _ = np.histogram(dataset4['z_circular_square'], bins=50, density=True)
plt.ylim(0, 350)
plt.xlabel("cosin simi(-1,1)")
plt.ylabel("Frequency")
plt.title('z circular&square cosin(-1,1)_1k')
mmd2_4 = mmd(hist2, hist4, sigma=1.0) # 计算MMD
kl2_4 = kl_divergence(hist2 / 1000, hist4 / 1000)
kl4_2 = kl_divergence(hist4 / 1000, hist2 / 1000)
count41 = np.count_nonzero((dataset4['z_circular_square'] >= 0.75) & (dataset4['z_circular_square'] <= 1))
count42 = np.count_nonzero((dataset4['z_circular_square'] >= 0.5) & (dataset4['z_circular_square'] <= 0.75))
count43 = np.count_nonzero((dataset4['z_circular_square'] >= 0.25) & (dataset4['z_circular_square'] <= 0.5))
count44 = np.count_nonzero((dataset4['z_circular_square'] >= 0) & (dataset4['z_circular_square'] <= 0.25))
count45 = np.count_nonzero((dataset4['z_circular_square'] >= -0.25) & (dataset4['z_circular_square'] <= 0))
count46 = np.count_nonzero((dataset4['z_circular_square'] >= -0.5) & (dataset4['z_circular_square'] <= -0.25))
count47 = np.count_nonzero((dataset4['z_circular_square'] >= -0.75) & (dataset4['z_circular_square'] <= -0.5))
count48 = np.count_nonzero((dataset4['z_circular_square'] >= -1) & (dataset4['z_circular_square'] <= -0.75))
print((mmd2_4, kl2_4, kl4_2))
print('图4' + str((count41, count42, count43, count44, count45, count46, count47, count48)))
# plt.text(x=-1, y=375, s='MMD:' + str(mmd2_4), fontdict=dict(fontsize=12, color='red', family='monospace', weight='normal'))
# plt.text(x=0, y=380, s='KL2_4:' + str(kl2_4), fontdict=dict(fontsize=12, color='red', family='monospace', weight='normal'))
# plt.text(x=0, y=365, s='KL4_2:' + str(kl4_2), fontdict=dict(fontsize=12, color='red', family='monospace', weight='normal'))
# plt.text(x=-1, y=300, s='cosin-[0.9,1] percentage:' + str(count41 / 1000), fontdict=dict(fontsize=12, color='red', family='monospace', weight='normal'))
# plt.text(x=-1, y=275, s='cosin-[-1,0.3] percentage:' + str(count42 / 1000), fontdict=dict(fontsize=12, color='red', family='monospace', weight='normal'))
plt.savefig('E:/FunctionMethod/analysis results/10.png')
plt.show()